It is only logical that cell-to-cell communications in a multicellular organism must be fast, and that they must be able to allow cells to respond to one another in diverse and complex ways. Typically, the intracellular signals used are molecules called “ligands,” and a given ligand can bind to a particular type of receptor on the surface of those cells that are to receive that signal. But this simple ligand binding alone is not enough to provide for the complex responses that the receiving cells may need to make. Cells therefore amplify and add complexity to this signal through complex, often cascading mechanisms leading to the rapid modulation of catalytic activities inside the cell, which in turn can produce complex, and sometimes dramatic, intracellular responses. This process as a whole, from initial ligand binding to completion of the intracellular response, is called “signal transduction.”
Signal transduction is often accomplished by the activation of intracellular enzymes that can act upon other enzymes and change their catalytic activity. This may lead to increases or decreases in the activity certain metabolic pathways, or may lead to even large intracellular changes, for example, the initiation of specific patterns of gene expression. The ability of one enzyme to alter the activity of other enzymes generally indicates that the enzyme is involved in cellular signal transduction.
The most common covalent modification used in signal transduction process is phosphorylation, which results in the alteration of the activity of those enzymes which become phosphorylated. This phosphorylation is catalyzed by enzymes known as protein kinases, which are often simply referred to as “kinases.”
Several key features of the kinases make them ideally suited as signaling proteins. One is that they often have overlapping target substrate specificities, which allows “cross-talk” among different signaling pathways, thus allowing for the integration of different signals (1). This is thought to be a result of the need for each kinase to phosphorylate several substrates before a response is elicited, which in turn provides for many types of diverse signaling outcomes. For example, a given kinase may in one instance transmit a growth inhibitory signal and in another instance transmit a growth promoting signal, depending on the structure of the extracellular ligand that has bound to the cell surface (2).
A second key feature is that the kinases are organized into several modular functional regions, or “domains” (3). One domain known as “SH3” is a proline-rich region of 55–70 amino acids in length, and another, known as “SH2,” is a phosphotyrosine-binding region of about 100 amino acids in length. These two domains are believed to be involved in recognizing and binding to the protein substrates. The third domain, “SH1,” is comprised of about 270 amino acids, and is the domain which is responsible for catalysis. It also contains the binding site for the nucleoside triphosphate which is used as energy source and phosphate donor (3). Other domains, including myristylation and palmitylation sites, along with SH2 and SH3, are responsible for assembling multiprotein complexes which guide the catalytic domain to the correct targets (3,22,23). Molecular recognition by the various domains has been studied using by x-ray diffraction and by using NMR methods (24–28).
These domains appear to have been mixed and matched through evolution to produce the large protein kinase “family.” As many as 1000 kinases are thought to be encoded in the mammalian genome (4), and over 250 kinases have already been identified. The large number of kinases and the large number of phosphorylation-modulated enzymes that are known to exist inside cells allow for rapid signal amplification and multiple points of regulation.
A third key feature of the kinases is their speed. The kinetics of phosphorylation and dephosphorylation is extremely rapid in many cells (on a millisecond time scale), providing for rapid responses and short recovery times, which in turn makes repeated signal transmission possible (5).
These features of the kinases have apparently led them to be used in a vast array of different intracellular signal transduction mechanisms. For example, growth factors, transcription factors, hormones, cell cycle regulatory proteins, and many other classes of cellular regulators utilize tyrosine kinases in their signaling cascades (12,13). Tyrosine kinases catalytically attach a phosphate to one or more tyrosine residues on their protein substrates. The tyrosine kinases include proteins with many diverse functions including the cell cycle control element c-abl (14–16), epidermal growth factor receptor which contains a cytoplasmic tyrosine kinase domain (12),c-src, a non-receptor tyrosine kinase involved in many immune cell functions (13), and Tyk2, a cytoplasmic tyrosine kinase which is involved in phosphorylation of the p91 protein which is translocated to the nucleus upon receptor stimulation and functions as a transcription factor (17). The serine/threonine kinases make up much if not all of the remainder of the kinase family; these catalytically phosphorylate serine and threonine residues in their protein substrates, and they have similarly diverse roles. They share homology in the 270 amino acid catalytic domain with tyrosine kinases. As such, although the discussion which follows focuses more particularly on the tyrosine kinases, that discussion is generally applicable to the serine/threonine kinases as well.
Unfortunately, the very features which make kinases so useful in signal transduction, and which has made them evolve to become central to almost every cellular function, also makes them extremely difficult, if not impossible, to study and understand. Their overlapping protein specificities, their structural and catalytic similarities, their large number, and their great speed make the specific identification of their in vivo protein substrates extremely difficult, if not impossible, using current genetic and biochemical techniques. This is today the main obstacle to deciphering the signaling cascades involved in tyrosine kinase-mediated signal transduction (4,6–8).
Efforts to dissect the involvement of specific tyrosine kinases in signal transduction cascades have been frustrated by their apparent lack of protein substrate specificity in vitro and in vivo (4,8). The catalytic domains of tyrosine kinases possess little or no inherent protein substrate specificity, as demonstrated by domain swapping experiments (18–23). The catalytic domain from one tyrosine kinase can be substituted into a different tyrosine kinase with little change in the protein substrate specificity of the latter (22).
The poor in vitro specificity of kinases also makes it difficult, if not impossible, to extrapolate what the in vivo function of given kinases might be. An isolated tyrosine kinase of interest will often phosphorylate many test protein substrates with equal efficiency (29). This apparently poor substrate specificity is also found in vivo; for example, many genetic approaches, such as gene knock out experiments, give no interpretable phenotype due to compensation by other cellular tyrosine kinases (30,31).
Another complication is that many tyrosine kinases have been proposed to phosphorylate downstream and upstream proteins which are themselves tyrosine kinases; although this appears to make complex positive feedback loops possible, it also makes dissecting the cascade even more difficult (1).
One important avenue for deciphering the role and understanding the function of enzymes, both in vitro and in vivo, is the use of specific enzyme inhibitors. If one or more compound can be found that will inhibit the enzyme, the inhibitor can be used to modulate the enzyme's activity, and the effects of that decrease can be observed. Such approaches have been instrumental in deciphering many of the pathways of intermediary metabolism, and have also been important in learning about enzyme kinetics and determining catalytic mechanisms.
In addition, such inhibitors are among the most important pharmaceutical compounds known. For example, asprin (acetylsalicylic acid) is such an inhibitor. It inhibits an enzyme that catalyzes the first step in prostaglandin synthesis, thus inhibiting the formation of prostaglandins, which are involved in producing pain (72). Traditional drug discovery can be characterized as the design and modification of compounds designed specifically to bind to and inactivate a disease-causing protein; the relative success of such an effort depends upon the selectivity of the drug for the target protein and its lack of inhibition of non-disease associated enzymes with similar enzyme activities.
Such approaches would appear to be promising ways to develop treatments for cancer, since many human cancers are caused by disregulation of a normal protein (e.g., when a proto-oncogene is converted to an oncogene through a gene translocation). And since kinases are key regulators, they have turned out to be very common proto-oncogenes, and thus ideal drug design targets.
The process of designing selective inhibitors is relatively simple in cases where few similar enzymes are present in the target organism, for example in cases where inhibitors of a protein unique to bacteria can be targeted. But unfortunately, the similarities between the kinases and their large number has almost completely frustrated the discovery and design of specific inhibitors, and has blocked most hopes of developing specific pharmaceutical treatments aimed at the proto-oncogene level. It is expected that the vast majority of candidate inhibitors will inhibit multiple kinases, even though they may have initially been identified as inhibiting a particular, purified kinase.
This is not to say, however, that inhibitors with at least some degree of kinase-specificity cannot be found. Several natural products have been identified which are relatively specific for particular kinase families, but attempts to derive general rules about kinase inhibition based on these has failed. Furthermore, as the following examples show, specificity in most cases is quite limited. For example, the compound Damnacanthal was reported to be a “highly potent, selective inhibitor” of the kinase p56lck (73); as shown in FIG. 2A, this compound has an inhibition constant (IC50) for that kinase which is almost seven times lower than for the kinase src (the IC50 is the concentration of inhibitor which must be added to reduce catalytic activity by 50%). The compound PPI (FIG. 2B) has a binding affinity for the kinase lck which is very strong (IC50=0.005 μM); but unfortunately, the inhibition of other kinases of the src family is very similar. It inhibits the kinase fyn with an almost identical IC50, 0.006 μM, and has only about a 4-fold higher IC50 for the kinase hck (IC50=0.020 μM). The compound CGP 57148 (FIG. 2C) has been reported to be “semi-selective” for the kinases abl (IC50=0.025 μM) and PDGFR (IC50=0.030 μM)(74). Nevertheless, considering the vast number of kinases and their relative cellular importance, and also considering that the above-described inhibitors have only been reported in the last two years, it appears that success in discovering or designing selective kinase inhibitors has been remarkably limited.
These difficulties described above have implications well beyond the mere frustration of scientists; they have frustrated efforts to decipher the kinase cascades and the function of individual kinases in those cascades and other cellular mechanisms. Such an understanding of kinase activity and function may be essential before certain human diseases can be effectively treated, prevented or cured. For example, it has been known for over 30 years that the oncogene bcr-abl is a protein kinase that is responsible for chronic myelogenous leukemia; but the physiological substrates that it acts upon to cause oncogenesis, which may be important drug design targets, have yet to be definitively identified (11). On the bright side, despite this shortcoming, the above-described inhibitor CGP 57148 is reportedly now undergoing clinical trials for use in treating myelogenous leukemia, even though the substrates it may block phosphorylation of in vivo are not known.
The medical significance of these difficulties is further illustrated by the Rous sarcoma virus (RSV), which has become an important model system for studying the role of kinases in oncogenesis. RSV transformation of fibroblasts is controlled by a single viral gene product, the protein tyrosine kinase v-src (32). It is the rapid time course and the dramatic morphological changes during RSV fibroblast transformation that have made RSV a paradigm for studies of oncogene activity in all cells. The origin (33), regulation (3,8,34,35), and structure (25,27,36) of v-Src have been extensively studied and are well understood (8,37,38). But central questions about this intensely studied kinase remains unanswered: what are its direct cellular substrates? Does inhibition of its catalytic activity effectively inhibit, or even reverse, transformation? Would such inhibition be an effective therapy for or prophylactic against RSV transformation? Unfortunately, as discussed above, the answers to these questions are not forthcoming, largely because the number of cellular kinases is enormous (it is estimated that 2% of the mammalian genome encodes protein kinases (4)) and because tyrosine kinases display overlapping substrate specificities (8,39) and share catalytic domains, making the design of specific inhibitors enormously difficult.
The expression of v-Src in fibroblasts results in the tyrosine phosphorylation of over 50 cellular proteins (37). These same substrates are also phosphorylated by other kinases in untransformed fibroblasts (40). Even the most sophisticated biochemical and genetic techniques, including anti-phosphotyrosine protein blots of transformed fibroblasts, transfection of fibroblasts with transformation-defective v-Src mutants, temperature-sensitive v-Src mutants, gene knock-out studies of cellular Src, host-range dependent Src mutants, anti-v-Src immunoprecipitation, and use of kinase-specific inhibitors, have not led to the unambiguous identification of v-Src's direct substrates (see reference (38) for a comprehensive review). But this situation is not unique; in fact, the direct substrates for the majority of cellular kinases remain unidentified (8). Furthermore, as discussed above, there also are remarkably few compounds known to selectively inhibit individual kinases, or even groups of related kinases.
Although the forgoing difficulties are daunting, new methods of rational drug design and combinational organic synthesis make the design or discovery of kinase-specific inhibitors feasible given sufficient resources. However, because the kinase networks are highly degenerate and interconnected in unknown ways, there is considerable uncertainty with regard to many diseases which kinases should be targeted for inhibition. Moreover, it is by no means clear that a specific inhibitor of a given kinase will have any effect on the disease, either in vitro or in vivo. Because kinases can be highly promiscuous, there is a significant chance that inhibiting one kinase will simply force another kinase to “take its place.” Therefore, there is a need for a simple and direct way to determine the biochemical and cellular effects of inhibiting a given kinase, before herculean efforts are undertaken to design or discover specific inhibitors.
From the forgoing, it is clear that there has been a long felt but unsatisfied need for ways to identify which cellular proteins are acted upon by individual protein kinases. Such a method would ideally also allow for the quantitative measurement of relative activity of a given kinase on its protein substrates, which could be used, for example, to detect how or whether actual or potential drug compounds might modulate kinase activity. In addition, there has also been a need for specific inhibitors of individual kinases or kinase families, which could be used to identify protein substrates (by looking for which proteins are not phosphorylated or are more weakly phospohorylated in the presence of the inhibitor), to study the biochemical and phenotypic effects of rapidly down-regulating a given kinase's activity, for use as drugs to treat kinase-mediated diseases, and to confirm that tedious efforts to design or develop more traditional inhibitor drugs would be worthwhile.
As is described in considerable detail below, the present invention for the first time provides a method for the highly specific inhibition of individual kinases, which have been engineered to bind the inhibitor more readily than the wild-type form of that kinase or other, non-engineered kinases. The invention also provides for the engineered kinases and the inhibitors to which they are adapted.
Moreover, as will become apparent, this method is even more broadly applicable, as it would provide similar advantages for the study of other enzymes which, like the kinases, covalently attach part of at least one substrate to at least one other substrate.
The present invention involves the engineering of kinases and other multi-substrate enzymes such that they can become bound by inhibitors which are not as readily bound by their wild-type forms. Modified substrates and mutant enzymes that can bind them have been used to study an elongation factor (41) and a receptor for cyclophilin A (42). However, prior to the present invention, it was not known how, or even if, multi-substrates enzymes which covalently attach part or all of a donor substrate onto a recipient substrate could be engineered to bind to an inhibitor, yet still retain at least some catalytic activity and at least some specificity for the recipient substrate in the absence of the inhibitor. The present invention is that this can be done, as explained below; and this invention for the first time opens the door to the selective inhibition of individual kinases, which are not only important tools for understanding of the kinase cascades and other complex catalytic cellular mechanisms, but also may provide avenues for therapeutic intervention in diseases where those mechanisms come into play.